91 results about "Bayesian neural networks" patented technology

The invention provides a criminal identification and forecast method. The method adopts a data pre-processing method in data mining; aiming at criminal information such as data, street address, criminal police zone, week, criminal type, criminal description and sentence processing, attribute reconstruction, feature extraction and feature selection are performed, the correlation between the criminal information is mined, a characteristic factor with maximum difference is generated, and the correlation between the characteristics factor and a criminal result, namely the criminal type is generated; and then a model integrating Gaussian Naive Bayes, a neural network, Logistic regression, regularized regression, K neighbor, random forest, a support vector machine and an XGBoost learning algorithm is built to obtain an element classifier based on a weighted voting classifier having highlight classification and favorable clustering effect, reconstructed data is analyzed, processed and identified, a criminal condition of a city in future is forecasted, an individual criminal map of the city is drawn, and the effects of promoting and regulating city public security and management are further achieved.

The invention relates to a semiconductor fabrication process multi-performance prediction method, comprising the steps of selecting an articles being processed level parameter, an equipment parameter and a workpiece parameter which represent the state of a semiconductor production line as influence factors of performance indexes; collecting relevant data of the production line, preprocessing by using a principal component analysis method, removing redundant information, constructing a multi-performance prediction model by using a Bayes neural network, and controlling the complexity of the network model by introducing a Bayes method; analyzing whether the model performance conforms to a performance evaluation criteria by a model precision proof method, and performing online correction on the network model structures which do not conform to the standard; finally determining the key factors influencing the average workpiece processing period and the equipment utilization rate. According to the semiconductor fabrication process multi-performance prediction method, the defects that the performance prediction model in the semiconductor field is limited by constraint conditions, the generalization performance is very poor and the like are improved, the problem that the single performance prediction model in the semiconductor field is not applicable to multi-performance prediction modeling is solved, and the method is an improvement of the semiconductor fabrication process multi-performance prediction method.

The invention discloses a wind power short-term power prediction method based on a relative error entropy evaluation method. The wind power short-term power prediction method comprises the following steps of: 1, acquiring historical data, and pre-treating the historical data to produce various training samples; 2, dynamically selecting the training samples, and establishing a bayesian neural network prediction model, an error feedback weighing time sequence prediction model and a wind power prediction unbiased grey verhulst prediction model; 3, respectively carrying out continuous prediction by adopting the three prediction models ten days ago from a prediction day; 4, respectively counting a relative error of each group of prediction data obtained in the step three, calculating an entropy and a variation degree coefficient of each group of relative error, and calculating a weight coefficient; 5, adopting the three prediction models to respectively carry out wind power prediction on the prediction day, and obtaining three groups of prediction data; and 6, carrying out combined prediction on the weight coefficient and the three groups of prediction data obtained in the step five to obtain a wind power short-term power prediction result. With the adoption of the wind power short-term power prediction method, the problem of determining the weight coefficient of combined prediction is solved, and the accuracy of wind power prediction can be improved.

The invention discloses a variation reasoning Bayesian neural network-based flood ensemble forecasting method. The method comprises the following steps of: setting dimensions of each layer of a Bayesian neural network; Selecting the prior probability distribution of the weight parameters of the Bayesian neural network, and parameterizing the weight parameters of the Bayesian neural network throughthe variational parameters to approximate the posterior probability distribution of the weight parameters of the Bayesian neural network; Calculating the relative entropy of the prior probability distribution and the variation posterior probability distribution, and calculating an expected log-likelihood function according to the training data set; Constructing an objective function according tothe relative entropy and the expected log-likelihood function; maximizing a target function, and training variational reasoning parameters; And carrying out ensemble forecasting on unknown flood by using the trained variational reasoning Bayesian neural network. According to the method, the variational reasoning is combined with the BNN model, and the posterior probability of the weight parametersof the Bayesian network model is approximated through variational distribution, so that the calculation process is simplified, the uncertainty of flood forecasting is quantitatively described, and the accuracy is improved.

The invention discloses a method and device for realizing a Bayesian neural network by using memristor intrinsic noise. The method comprises the steps: obtaining a Bayesian network, and carrying out the training of the Bayesian network according to a selected data set, and obtaining the weight distribution of the Bayesian network; and processing the weight distribution of the Bayesian network, calculating according to the processed weight distribution and the conductivities of the plurality of memristors to obtain a target conductivity value, and mapping the target conductivity value into thememristors. According to the method, the Bayesian neural network is realized by utilizing the memristor cross array, and the power consumption is low, and the calculation speed is high, and the calculation energy efficiency is high.

The invention discloses a steam pipe network prediction system based on a Bayesian neural network algorithm. The steam pipe network prediction system based on the Bayesian neural network algorithm mainly comprises a relational data base, a data collection module, a data display module and a Bayesian neural network prediction module. The data display module is arranged on an engineer station, the Bayesian neural network prediction module is arranged on an application server, the relational data base is arranged on a relational data base server, and the data collection module is arranged on a real-time data base. The relational data base is a data communication medium between the data display module and the Bayesian neural network prediction module, the Bayesian neural network prediction module writes results of the Bayesian neural network prediction module into the relational data base, and the data display module reads and displays the results from the relational data base. The steam pipe network prediction system based on the Bayesian neural network algorithm has the advantages of achieving rapid solution, ensuring precision of a calculation model and being capable of meeting requirements of a process technology, and the calculation results are in fit with actual operation conditions.

The invention discloses a power grid energy management method and system based on a double-depth expectation Q-learning network algorithm. The method comprises the following steps: firstly, modeling photovoltaic output uncertainty of a prediction point based on a Bayesian neural network, and obtaining probability distribution of photovoltaic output; inputting the probability distribution of the photovoltaic output into a power grid energy management model based on a double-depth expectation Qlearning network algorithm to obtain a corresponding photovoltaic power generation output strategy; and enabling the system to operate each photovoltaic output device for application according to the photovoltaic power generation output strategy; according to the invention, a microgrid economic dispatching problem is simulated into a Markov decision process, an objective function and constraint conditions are mapped into a reward and punishment function for reinforcement learning, and an optimal decision is obtained by using learning and environment interaction capabilities of the reward and punishment function; the Bayesian neural network is used to model the uncertainty of the photovoltaic power generation output in the learning environment, and the state random transfer is properly considered in the Markov decision process, and therefore the convergence rate of the algorithm is significantly improved.

A trained model is described in the form of a Bayesian neural network (BNN) which provides a quantification of its inference uncertainty during use and which is trained using marginal likelihood maximization. A Probably Approximately Correct (PAC) bound may be used in the training to incorporate prior knowledge and to improve training stability even when the network architecture is deep.

The invention discloses a frequency division duplex downlink transmission method based on Bayesian neural network channel prediction. The method comprises the following steps: (1) storing a pluralityof pieces of past channel state information as training data by a base station, and training a Bayesian neural network with initialized parameters by utilizing the training data; (2) in the downlink transmission process, the base station predicting the channel state information of the next transmission time slot by using the trained Bayesian neural network; at the next transmission time slot, thebase station using the predicted channel state information for precoding and transmission of data to be sent in an uplink feedback waiting stage; (3) the base station obtaining estimated channel stateinformation through the fed back information, and sending downlink data in a transmission stage by utilizing the estimated channel state information; and meanwhile, the base station continuously predicting the channel state information of the next transmission time slot through the Bayesian neural network. According to the invention, the downlink transmission rate of the large-scale MIMO system can be further improved.

The invention discloses a heat accumulation electric boiler and clean energy prediction matching and consumption control method, and the method comprises the steps: collecting weather factor data and photovoltaic energy curtailment data, and obtaining a training sample set through normalization processing; 2, designing a fuzzy Bayesian neural network model comprising an input layer, a hidden layer and an output layer, and selecting an excitation function, a training function, and a learning function; 3, applying the obtained optimal network prediction model in a distributed photovoltaic power generation system, so as to obtain the photovoltaic energy curtailment quantity of a photovoltaic power station under the different weather factor conditions; 4, giving consideration to an economic performance index under the condition that the photovoltaic energy curtailment quantity of the photovoltaic power station is predicted, and enabling a distributed heat accumulation electric boiler to extremely consume the photovoltaic energy curtailment quantity through the reference of the predicted photovoltaic energy curtailment quantity and an index which enables the combined operation benefit of the photovoltaic power station and the heat accumulation electric boiler and environment benefit. The method solves problems that a conventional consumption mode is not flexible, is low in consumption efficiency, and is not sufficient in consumption capability.

The invention relates to a method for predicting stratospheric airship skin material deformation through a neural network, and belongs to the technical field of damage analysis. The implementation method comprises the following steps: aiming at the true deformation of a stratospheric airship skin material in a complex working environment close to non-proportional biaxial tension, carrying out biaxial tension tests under various stress ratio conditions, and collecting training sample data required by a neural network; constructing a Bayesian neural network comprising an input layer, a hidden layer and an output layer so as to establish a deformation behavior simple expression model of the skin material; and adopting the trained neural network to predict the deformation behavior of the skinmaterial in real time. The method for predicting the stratospheric airship skin material deformation by using the neural network provided by the invention is relatively high in prediction precision, relatively good in stability and strong in popularization capability, can meet the requirement of accurately predicting the skin material deformation behavior in real time, and provides a new method for optimizing the material design and guiding the stratospheric airship structure design.

The invention relates to an AGC system identification method based on big data and a Bayesian neural network. The method includes: 1, collecting historical data of an AGC system, and obtaining sampledata after preprocessing; 2, initializing a Bayesian neural network; 3, calculating that input and output of each neuron in a hidden layer and an output layer, calculating the difference between the actual output and the output of the Bayesian neural network, and calculating the error according to the MSE standard; 4, judging whether that error meet the requirement or not, if so, carrying out step6, otherwise, correcting the weights and threshold between the output layer and the hidden layer, correcting the weights and thresholds between the input layer and the hidden layer, updating the connection weights, and increasing the learning times by 1; 5, repeating that steps 3 to 4 until the error requirement or the maximum learn times are met; 6, computing the Bayesian neural network and obtaining a mathematical model of identification. Compared with a classic BP neural network identification method, the method has better identification accuracy and faster convergence speed.

The invention relates to a streaming media code rate self-adaption method and device supporting a neural network and computer equipment. The method comprises the following steps: acquiring a historical network throughput measurement value, a preset vector of available resolution and current buffer occupation information, inputting the historical network throughput measurement value, the preset vector of available resolution and the current buffer occupation information into a pre-constructed Bayesian neural network, outputting a throughput prediction value of a next time period, constructing a model prediction control optimization model by taking preset QoE index optimization as a target, and solving to obtain a predicted downloading bit rate of the current video block; after execution, obtaining a corresponding reward value according to the QoE index, continuously training the Bayesian neural network according to the predicted downloading bit rate and the reward value, and adaptively obtaining the optimal bit rate of the downloaded video block according to the continuously trained Bayesian neural network and the model prediction control optimization model. The throughput prediction accuracy and the mobile network video quality are improved.

The invention relates to the field of intelligent analysis of monitoring data signals, in particular to an intelligent analysis system for a substation equipment monitoring data signal, and aims to provide the intelligent analysis system for converting data information into an event notification mode from a conventional item-by-item notification mode and establishing a corresponding relationship and an internal relationship between notification information and monitored equipment. According to the intelligent analysis system provided by the invention, the monitoring data signal collected by amonitoring system is learnt and analyzed through an intelligent learning algorithm of a Bayesian neural network, thereby assisting a monitoring manager to analyze and manage the monitoring signal; andaccording to established intelligent monitoring information processing strategy library and network topology structure library, in combination with the specific monitoring signal, automatic perception, analysis and decision making during sending of warning information are realized, so that the pressure of working personnel is reduced, a large amount of repeated and complex workloads of the working personnel are reduced, and an efficient and reliable management mode is provided for monitoring management.

The invention discloses a photovoltaic probability prediction method and system based on a Bayesian neural network. The method comprises the steps of obtaining weather forecast data of a to-be-predicted point and historical output data of photovoltaic equipment; carrying out dimension reduction processing on the weather forecast data, and obtaining feature data based on the weather forecast data after dimension reduction processing and historical output data of photovoltaic equipment; and substituting the feature data into a pre-constructed improved Bayesian neural network model to obtain photovoltaic output distribution of the to-be-predicted point. According to the method, the photovoltaic output distribution of the to-be-predicted point is obtained, compared with a deterministic prediction mode, the photovoltaic probability prediction method provided by the invention has a smaller average interval width when the same prediction accuracy is achieved, the prediction precision is improved, and the method has important significance for improving the safety and stability of a power grid.

The invention relates to an inference method, in particular to a Bayesian network inference method based on a random computing theory. According to the method, the calculation process of the Gaussianrandom generator based on the central limit theorem is simplified; the multiplication operation in the Bayesian neural network reasoning process is simplified; a 0-1 sequence obtained by using a binomial distribution generator is used as a data unit, multiplication operation is realized by using an AND gate, and a 0-1 random number sequence directly participates in calculation to complete Bayesianneural network inference, so that the purposes of reducing hardware resources, improving the calculation speed and reducing the system power consumption are achieved.

The invention belongs to the technical field of image recognition, and discloses a specific character recognition FPGA implementation method and system, a storage medium and application, which are used for detecting the matching degree of a one-dimensional sequence and a plurality of feature sequences, and are expressed in two dimensions as follows: characters in a specific character set are recognized; based on a Bayesian neural network (BNN) and a random calculation theory: the voting system comprises a voting result statistics module, a multi-input comparison module, a multi-voter voting module, a single-voter voting module, a pixel flow matching specific feature module, a 1 counting module and a random sequence generation module. Aiming at an MNIST data set training result, the methodcomprises the following implementation steps: converting input handwritten numbers, converting each pixel into a random calculation theoretical number with a 128bit bit width, sequentially inputting input streams, and obtaining an identification result after a fixed time period after the input is finished. The method has the advantages of being high in recognition speed, high in accuracy, suitablefor hardware implementation, relatively low in resource consumption and expandable in application range.

The invention discloses a medical image classification method and system and a storage medium, which can be applied to the technical field of image classification. The method comprises the following steps: respectively segmenting sequence images, and constructing a target area three-dimensional image corresponding to a target area image obtained by segmentation; inputting the three-dimensional image of the target area into a full convolutional neural network model to obtain an image disease probability graph; inputting the image disease probability graph into a Bayesian neural network model to obtain a classification result and uncertainty corresponding to the medical image; generating a fitting curve of credibility and uncertainty intervals according to the classification results and uncertainty corresponding to all the medical images; when the fitting curve meets a preset requirement, determining an uncertain target interval; and determining that the uncertainty corresponding to the medical image belongs to an uncertainty target interval, and taking a classification result corresponding to the medical image as a target classification result. According to the invention, the classification result of the medical image given by the current classification algorithm can better conform to the actual situation.

The invention relates to a tunnel surrounding rock geological classification information prediction method based on a Bayesian neural network, and the method comprises the steps: collecting surrounding rock geological classification information of an existing tunnel and a fine collection tunnel under construction, carrying out the normalization processing, and determining the probability distribution of the tunnel surrounding rock geological classification information through Monte Carlo random analysis; preliminarily determining the number of nodes of an input layer, a hidden layer and an output layer of the Bayesian neural network model so as to establish a Bayesian neural network prediction model by utilizing the existing tunnel engineering data with similar geological information; andupdating the prediction model in real time by utilizing tunnel surrounding rock geological classification information newly obtained in the excavation process along with continuous forward advancing of the working face, and further gradually improving the prediction precision of the model. The prediction method provided by the invention has good universality and high prediction precision, can makeeffective judgment on geological classification information of an unknown section in front of tunnel excavation in advance, and is suitable for prediction of geological classification information ofmost tunnel surrounding rocks.

The present invention discloses a neural network training method for a memristor memory for memristor errors, which is mainly used for solving the problem of decrease in inference accuracy of a neural network based on the memristor memory due to a process error and a dynamic error. The method comprises the following steps: performing modeling on a conductance value of a memristor under the influence of the process error and the dynamic error, and performing conversion to obtain a distribution of corresponding neural network weights; constructing a prior distribution of the weights by using the weight distribution obtained after modeling, and performing Bayesian neural network training based on variational inference to obtain a variational posterior distribution of the weights; and converting a mean value of the variational posterior of the weights into a target conductance value of the memristor memory.

The invention discloses a lithium battery temperature estimation method and system based on a Bayesian neural network. The method comprises the following steps: collecting battery electrochemical impedance spectroscopy data and a temperature label; processing the electrochemical impedance spectroscopy data of the battery based on an ARD algorithm to obtain temperature-dependent characteristics and temperature-dependent impedance frequency points; training a Bayesian neural network model based on the temperature-related features and the temperature labels, and collecting impedance data under the temperature-related impedance frequency points; and inputting the impedance imaginary part data into the temperature estimation model to obtain the internal estimated temperature and confidence interval of the battery at the current moment. The system comprises an offline data acquisition module, a temperature related data determination module, a model training module, an online data acquisition module and a temperature estimation module. According to the invention, accurate internal temperature estimation of the whole life cycle of the power battery is realized. The lithium battery temperature estimation method and system based on the Bayesian neural network can be widely applied to the field of battery thermal management.

The invention relates to a point cloud recognition and segmentation method based on a Bayesian neural network, and the method builds the Bayesian neural network, and comprises three parts: feature extraction, recognition and segmentation. Wherein the feature extraction part comprises three Bayesian convolution layers, and an activation layer is connected behind each convolution layer. The networkstructure of the identification part is a three-layer full-connection layer, and the number of nodes of the final output layer is the same as the number of categories. Wherein the network structure ofthe segmentation part is divided into three Bayesian convolution layers, an active layer is connected behind each convolution layer, the number of nodes of the final output layer is equal to the number of points contained in the point cloud, and each node outputs a vector of which the length is equal to the number of categories. The point cloud data firstly pass through the feature extraction part to obtain feature values, and then the feature values are respectively input into the recognition part and the segmentation part to obtain recognition and segmentation results.

The invention belongs to the technical field of power load prediction, and specifically relates to a power load prediction method. The power load prediction method comprises the following steps: S1, constructing a training data set and a test data set of a prediction model; S2, establishing a prediction model based on a hierarchical parallel Bayesian neural network; S3, inputting training samplesin the training data set into the prediction model based on the hierarchical parallel Bayesian neural network for training; and S4, inputting test data into a trained prediction model for prediction so as to obtain a power load prediction result. According to the invention, the power load is predicted by using the prediction model of the hierarchical parallel Bayesian neural network; the practicability of the model provided by the invention is verified by using actual power grid load data; and high prediction precision is achieved.

The invention discloses a marketing prediction method combining inner/outer product feature interaction and a Bayesian neural network. The method comprises a data preprocessing step, a data set division step, a model establishment step and a marketing activity click prediction step. According to the method, in the building process of a prediction model, the prediction uncertainty is introduced into the Bayesian neural network by effectively utilizing Bayesian inference, so that the Bayesian neural network model has higher robustness. And by adopting an inner/outer product combination method, the features are crossed to extract high-dimensional recessive features. Therefore, the application of deep learning to advertisement calculation and recommendation system algorithm problems can be effectively expanded, and the accuracy of user click behavior prediction is remarkably improved.